Optimized 3 T EPI of the amygdalae
Introduction
The amygdalae are approximately 1.7 cm3 nuclei (Brierley et al., 2002) in the caudal brain responsible for interpreting the emotional content of sensory and cognitive information (Zald, 2003), and are therefore important in the normal function of memory, conditioning, motivation, arousal, and attention (Phan et al., 2002), as well as disorders with affective components such as depression (Posse et al., 2003), autism (Howard et al., 2000), psychopathia (Patrick, 1994), and schizophrenia (Gur et al., 2002). Functional fMRI in the caudal brain using GR-EPI is problematic due to large magnetic susceptibility differences that arise at the interfaces between bone, air, and tissue, primarily in the auditory canals and sphenoid sinus. These cause inhomogeneity in the main magnetic field which leads to intravoxel dephasing and signal loss, and mislocalization of signal (image distortion). A frustration to the attempt to draw group functional conclusions is that measured amygdalae volumes vary almost 4-fold between studies (Brierley et al., 2002), reflecting a genuinely wide range of volumes as well as of measurement methods, and location is heterogeneous between subjects, making normalization to digital brain atlases, which are usually based on a single subject's anatomy (Toga and Thompson, 2001), problematic. Draining veins may also cause erroneous activation-like signals from macroscopic vessels Frahm et al., 1994, Lai et al., 1993 near the amygdalae as poststimulus hyper-oxygenated blood is transported from distant activated regions.
Technical strategies for reducing susceptibility-related signal loss include slice-select gradient compensation (Merboldt et al., 2000), tailored RF pulses (Cho and Ro, 1992), refined shimming methods (e.g., z-shimming, Deichmann et al., 2002, Glover, 1999), and the use of short effective echo times (TEeff), high resolution, and thin slices (Young et al., 1988). The effectiveness of these has recently been reviewed (Hennig et al., 2003). Geometric correction based on a map of local deviations from the main magnetic field promises an elegant solution to distortion, but a field map from a single time instance may be inadequate if motion is substantial, particularly in high-resolution studies. Regions of high-susceptibility gradients are also particularly likely to cause the appearance of activation if there is a correlation between the stimulus and head motion Hajnal et al., 1994, Hutton et al., 2002.
The number of neuroscience and clinical studies concentrating on the amygdalae has accelerated during the last years, and these have not, in the main, availed themselves of technical advances designed to counter the problems posed in this area. All studies to date have used large voxel sizes, and none have shown original echo planar images illustrating signal in the amygdala in publications or verified voxel-by-voxel activation. Some investigators in the field have questioned the veracity of reports of activation in the amygdalae (Merboldt et al., 2001), showing that, with standard parameters and at field strengths of 2 T or less, little or no signal is achieved in the amygdalae in echo planar images.
The ideal parameters for single-shot GR-EPI-based functional studies of the limbic system are assessed here via an examination of magnetic field variations in the region (leading to the ideal slice orientation and other gradient directions), and a systematic examination of the influence of resolution, slice thickness, and effective echo time on SNR in the amygdalae in single repetition EPI data of healthy subjects. The optimized protocol is also tested against a standard protocol via voxel-wise SNR achieved in a time-series study, which, including physiological fluctuation, correlates better with BOLD sensitivity than SNR in single repetition images (LaBar et al., 2001). The vascular context of the region has been examined using high-resolution BOLD venography (Reichenbach et al., 2000).
Section snippets
Methods
Seventeen healthy subjects (13 female, 4 male, in the age range of 21–54 years), from whom written informed consent was obtained, were scanned with a 3 T Medspec S300 whole body system (Bruker Biospin, Ettlingen, Germany) with whole body gradients (BGA-55 max. 45 mT/M in <300 μs) and standard birdcage head resonator. In all cases, global linear shims only were performed to evoke conditions typical in fMRI studies. A number of single-shot, single repetition GR-EPI protocols and other scans
Data analysis
FLASH scans with echo times of 6.0, 6.5, and 7.5 ms (“Scan 1”, “Scan 2”, and “Scan 3”, with all parameters listed in Methods, Item 1) were reconstructed to phase and magnitude images. Regions of brain were segmented using Brain Extraction Tool (BET, (Smith, 2002)), and these were used to mask phase images, which were unwrapped in the temporal domain (Windischberger et al., 2003) using in-house developed software implemented in Interactive Data Language (IDL, Research Systems, Inc., Boulder,
Results
Deviations in magnetic field across the amygdalae are listed individually for the six subjects and for the group in Table 1. B0 values were higher caudally, anteriorly, and medially, as can be seen in Fig. 1, with the exception of subject 3, right amygdala, ML direction. Variations were approximately linear in the ML, AP, and CC directions. Linear fits to average field deviations in these three direction yielded average gradients over all subjects of 1.44 ± 0.52, 1.46 ± 0.52, and 1.0 ± 1.3
Discussion
The optimum imaging parameters for fMRI studies focussing on the amygdalae and based on 3 T single-shot GR-EPI are established here, with relevance also to other field strengths.
Chen et al. (2003) measured main magnetic field variations across the amygdalae in three subjects at 3 T to be highest craniocaudally (a factor of approximately 6 times greater than in AP and almost 100 than in ML, albeit with large intersubject variability), and suggest subject-dependant gradient orientation. We find
Conclusion
Susceptibility-related signal loss in the amygdala region could be substantially mitigated using the optimized high-resolution parameters suggested here. The reduction in brain coverage or temporal resolution required to achieve this with current hardware implies that fMRI studies of functional networks which span the brain (e.g., regions responsible for both emotion and cognitive processing) may, for the moment, be unviable.
A baseline study with optimized voxel sizes of approximately 6 μl (128
Acknowledgements
S. R. gratefully acknowledges receipt of a Human Frontiers Science Program short-term fellowship award (ST00004/2002-C/1). C. W. and S. R. were also supported via the Austrian Science Fund (P16669-B02 to E. M.), and C. W. and A. R. via the Jubiläumsfonds of the Austrian National Bank (P9305 to E. M.). The anonymous reviewers are thanked for constructive criticism and helpful suggestions.
References (33)
- et al.
The human amygdala: a systematic review and meta-analysis of volumetric magnetic resonance imaging
Brain Res. Brain Res. Rev.
(2002) - et al.
Selection of voxel size and slice orientation for fMRI in the presence of susceptibility field gradients: application to imaging of the amygdala
NeuroImage
(2003) - et al.
An evaluation of the use of magnetic field maps to undistort echo-planar images
NeuroImage
(2003) - et al.
Compensation of susceptibility-induced BOLD sensitivity losses in echo-planar fMRI imaging
NeuroImage
(2002) - et al.
Echo time dependence of BOLD contrast and susceptibility artifacts
NeuroImage
(2002) - et al.
Image distortion correction in fMRI: a quantitative evaluation
NeuroImage
(2002) - et al.
Comparison of multi-echo spiral and echo planar imaging in functional mri
Magn. Reson. Imaging
(2002) - et al.
Reducing inhomogeneity artifacts in functional MRI of human brain activation—thin sections vs gradient compensation
J. Magn. Reson.
(2000) - et al.
Functional MRI of the human amygdala?
NeuroImage
(2001) - et al.
Functional neuroanatomy of emotion: a meta-analysis of emotion activation studies in pet and fmri
NeuroImage
(2002)